Genomic analysis establishes correlation between growth and laryngeal neuropathy in Thoroughbreds

Equine recurrent laryngeal neuropathy (also known as idiopathic laryngeal hemiplegia) is a common upper respiratory tract disorder of horses and a significant cause of their reduced athletic performance [1]. RLN represents a distal axonopathy of the recurrent laryngeal nerves, which is clinically expressed only on the left side of the larynx. Interestingly, the left nerve is the longest nerve in horses [2, 3]. RLN is also a mononeuropathy, as other peripheral nerves of the horse remain unaffected [4]. As a result of the axonal degeneration, secondary paresis to complete paralysis of the intrinsic laryngeal abductor and adductor muscles ensues. Although RLN results in fiber-type grouping in all of the intrinsic laryngeal muscles, it is neurogenic atrophy of one particular abductor, the Musculus cricoarytenoideus dorsalis (CAD), that prevents maximal abduction of the arytenoid cartilage and contributes to the development of airway obstruction. The degree of CAD paresis/paralysis is graded endoscopically from 1 to 4 [5] and a significant correlation between the laryngeal grade and the histopathological changes of the CAD has been documented [6].

Recurrent laryngeal neuropathy is an important axonopathy for horses in competitive events, because the resultant paresis of the intrinsic laryngeal muscles leads to obstruction of air flow during intense exercise, an abnormal inspiratory noise known as “roaring,” and, most importantly, reduced performance. Despite its importance, however, the etiology of the axonopathy is unknown. RLN has particular economic importance in Thoroughbred racehorses because these animals enter race training as 2-year-olds. Even though all horses going through sales are endoscopically examined for evidence of RLN, the condition may not yet be clinically evident in some animals [7] so a diagnostic test to predict RLN risk would be very useful.

The prevalence of RLN varies, ranging from between 2-11% in Thoroughbreds [8, 9] and 35-46% in the Draft horse breeds [10, 11]. In draft breeds, taller horses are at greater risk of developing RLN, but the relationship between height and RLN in Thoroughbreds has not been clearly defined [11, 12]. From a genetic point of view, all modern Thoroughbreds originated from three stallions imported into England in the 17^th and 18^th century that were crossbred to local mares [13]. As a consequence of this restricted founder stock and strong selection for racing performance, modern Thoroughbreds exhibit low genetic diversity and high linkage disequilibrium [14], making them particularly well suited to identify genetic associations using a genome-wide association study (GWAS) approach.

The presence of genetic factors contributing to the pathogenesis of RLN is suggested by the observation that offspring of RLN-affected stallions are more likely to be affected with the disorder than are offspring from unaffected stallions [15, 16]. Elucidating the genetic mechanism causing the disease is of particular research interest since, unlike most myelinopathies, RLN manifests itself as a bilateral mononeuropathy [4, 17]. However, the genetics of RLN have not been extensively investigated. A GWAS of RLN has been conducted [17] in a mixed population of horses consisting of 234 cases (196 Warmbloods, 20 Trotters, 14 Thoroughbreds, 4 Draft horses) and 228 breed-matched controls. Using a haplotype-based approach based on Illumina Equine SNP50 genotypes, the investigators identified two protective loci in the Warmblood horses on chromosomes 21 (P = 1.62 × 10^-6) and 31 (P = 1.69 × 10^-5) associated with RLN. Neither of these chromosome regions contained regions associated with neuropathies in other mammals or candidate genes that could be immediately linked to peripheral neuropathy. The present study reports on use of a newer Illumina Equine SNP70 BeadChip to conduct a GWAS on a large cohort of Thoroughbred horses that were phenotypically characterized for RLN and height. Specifically, we sought to identify risk loci associated with RLN and in doing so, gain additional insight into the etiopathogenesis of this disease. Because of the clinically reported connection between height and RLN, we also examined their genetic association.

The selective breeding of domestic animals for desirable traits has often been accompanied by the appearance of genetically determined undesirable characteristics: examples include malignant hyperthermia in swine selected for rapid increase in muscle mass, osteoporosis in chickens selected for high rates of egg production, and hyperkalemic periodic paralysis in Quarterhorses selected for muscular appearance. In general, horses have been selectively bred for power and speed. The former requires size and muscle mass while the latter requires height (long legs), a compact muscle mass close to the trunk, and high rates of maximal oxygen consumption. Although selection for height apparently has been accompanied by an increased prevalence of RLN, the genetic basis for this is currently unknown.

The current study, which used the largest cohort of RLN-affected and unaffected horses ever assembled from a single breed revealed a very strong genetic correlation between RLN and height. This association was previously well known across breeds (ponies are rarely clinically affected by RLN and draft horses frequently are), but has never been demonstrated over the small range of heights (152 to 178 cm) spanned in the present study. This was evident not only in the clinical observations (cases were taller than controls), but also in the GWAS results. The haplotype associations for height and RLN were coincident and the genetic correlation between the traits was >90%.

The phenotype of RLN was defined based on the presence (cases) or absence (controls) of endoscopically visible failure of abduction of the left arytenoid and associated vocal fold. This clinical and functional definition was used because veterinarians in the field use it and because it has important implications for athletic performance. There were two reasons for selection of grade 3 as the threshold score for a case: first, horses with this grade are unable to maintain vocal fold abduction and therefore require surgical correction to restore their performance; and second, failure of abduction is easily recognized endoscopically by experienced veterinarians (those participating in selection of cases examine many racehorses before sale as well as clinical cases). However, histopathological examination of the intrinsic muscles of the larynx can reveal muscle fiber-type regrouping (7evidence of axon death and reinnervation from a neighboring axon) in some horses that do not have clinical evidence of laryngeal paresis [28] and therefore would be controls under the definition used in the present investigation. Furthermore, foals can demonstrate fiber-type regrouping at birth [29] and ponies have fiber-type regrouping but rarely develop the clinical syndrome. Definition of cases and controls based on the presence or absence of fiber-type regrouping is not currently practical, and it is possible that clinical expression of RLN involves factors other than just changes in muscle fiber type.

Recently, several independent studies have documented strong associations between the LCORL/NCAPG region and body size in the horse [30, 31], but no conclusive causative mutation has been detected to date. A role for both LCORL and NCAPG in growth is supported by well-validated associations to human height [32, 33], and to fetal growth rate in cattle [34].

Within the Thoroughbred breed, the ECA3 locus appears to have the single largest effect on withers height (Figure 1B). Previous work in 16 diverse breeds selected for extreme body size identified three significant loci in addition to ECA3 (HMGA2, ZFAT, and LASP1) as well as a Thoroughbred-specific QTL on ECA28, but no signal for these loci was detected in this dataset [26]. Although marker coverage for HMGA2 is poor on the Illumina SNP70 product, the same markers for ZFAT are included in this analysis and segregate at ~ 20% in our cohort, but do not appear to impact withers height. The association of ZFAT with height is replicated in a study of 1,077 Franches-Montagnes horses [25], suggesting the effect of the size QTL near ZFAT may be dependent on genetic background. The previous estimate that 59% of the variation in height in Thoroughbreds was explained by LCORL/NCAPG and the QTL on ECA28 is significantly higher than our estimate of 24% based on the entire SNP dataset. This may reflect an overestimation in their analysis due to small sample size (N = 48) or an underestimation in ours due to the small range of Thoroughbred heights, environmental variability, and/or measurement error.

Our study confirms anecdotal evidence and previous reports in draft breeds identifying an association between RLN and large size in horses [11, 12]. However, the nature of this association remains unknown. Certainly, large body size will increase the overall length of the recurrent laryngeal nerve, perhaps increasing the risk for mechanical injury of the nerve during athletic performance or for damage associated with defects in axonal transport. Alternatively, LCORL and NCAPG are genes encoding transcription factors and could have diverse targets with varying physiological functions. Thus, the same causative mutation may be implicated for both body size and RLN phenotypes. Lastly, given extensive LD across the ECA3 locus, RLN and body size may be influenced by two independent genetic changes associated by linkage. Further work will be needed to distinguish which of these three hypotheses is the most likely explanation for the association between this ECA3 locus and RLN.

Besides the LCORL/NCAPG locus, the chromosome 18 and X candidate loci are near to genes that are likely to affect muscle physiology and growth in horses (MSTN and ITM2A). Interestingly, the primary association signal in [17] is also located in a myosin gene (MYO9B), a candidate gene that may be very relevant to the pathology of a distal axonopathy like RLN. Yet, our study suggests that QTLs on chromosome 21 and 31 identified by Dupuis et al.[17] are not an important influence on RLN in our Thoroughbred cohort. Genetic background may play an important role in a polygenic and multifactorial condition like RLN. Environmental conditions are also drastically different in the actively training racehorse (the majority of animals sampled for this work) and Warmblood horses used primarily for the sports of show jumping and dressage. Mean age of animals between these two studies was also quite different. Cases used here were on average just 3 years of age versus ~8 years of age in the Dupuis study ([17]; Dupuis personal communication). Therefore, the ECA3 locus may be most important for development of RLN in young horses in training for near maximal athletic performance.

The results of this study add further confirmation to the association between height and the clinical occurrence of RLN and indicate the close relationship between the genetic loci determining these two characteristics. However, despite being statistically significant, LCORL/NCAPG accounted for only 6% of the variation in RLN. However, in these Thoroughbreds, the proportion of variation of RLN explained by the entire SNP70 set (“chip heritability”) was 45%, which, although much higher than in most complex diseases [35], suggests that multimarker tests will be needed for accurate prediction of RLN risk. Because of the nearly complete genetic correlation between height and RLN risk in Thoroughbreds, it may be difficult reduce RLN prevalence without altering body size or other growth characteristics that may be important for racing performance. However, our results do not preclude the possibility of modifier loci that attenuate RLN risk without reducing size or performance, or that the RLN risk allele is distinct but tightly linked to the body size locus on ECA3.